Hyperion2: A Toolkit for {Meta-, Hyper-} Heuristic Research

Citation

Brownlee A, Swan J, Ozcan E & Parkes AJ (2014) Hyperion2: A Toolkit for {Meta-, Hyper-} Heuristic Research. In: Proceedings of the 2014 Conference Companion on Genetic and Evolutionary Computation Companion. GECCO Comp '14. GECCO 2014: Genetic and Evolutionary Computation Conference, Vancouver, BC, Canada, 12.07.2014-16.07.2014. New York, NY, USA: ACM, pp. 1133-1140. http://doi.acm.org/10.1145/2598394.2605687; https://doi.org/10.1145/2598394.2605687

Abstract
In order for appropriate meta-heuristics to be chosen and tuned for specific problems, it is critical that we better understand the problems themselves and how algorithms solve them. This is particularly important as we seek to automate the process of choosing and tuning algorithms and their operators via hyper-heuristics. If meta-heuristics are viewed as sampling algorithms, they can be classified by the trajectory taken through the search space. We term this trajectory a trace. In this paper, we present Hyperion2, a JavaTM framework for meta- and hyper- heuristics which allows the analysis of the trace taken by an algorithm and its constituent components through the search space. Built with the principles of interoperability, generality and efficiency, we intend that this framework will be a useful aid to scientific research in this domain.

Keywords
analysis; experimental framework; hyper-heuristics; metaheuristics; search space